ci: build releases on the dedicated runner pool #1
Reference in New Issue
Block a user
Delete Branch "ci/dedicated-runners"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Contexto
El release de zigfetch pedía
ubuntu-latest, que hoy atienden los runners de sitio de infra01. Es uno de los consumidores que faltaban en la migración de bitua/network#243.Cambio
runs-on: [x86_64, debian]. El job publica un assetzigfetch-x86_64-linux.zip, así que queda pineado a amd64 en vez del pool mixto.La imagen
runner-image:trixietrae sudo, apt, zip, curl, jq y node 24 para las JS actions (actions/checkout@v4,mlugg/setup-zig@v2); el step delibpci-devsigue instalándose por apt como antes. No cambian versiones ni el flujo de release.Validación
YAML validado localmente. El workflow sólo dispara en tags
v*, así que la prueba real llega con el próximo release; no se creó un tag de prueba para forzarla.✅ Botua — Approved
Summary
This PR migrates the
zigfetchrelease workflow offubuntu-latestand onto the dedicated self-hosted runner pool by changing a single line in.gitea/workflows/release.yml:13toruns-on: [x86_64, debian]. The change is small, well-scoped, and consistent with the broader runner migration (the siblingbitua/snitchrelease workflow uses the exact same label set). Since the job publishes a nativezigfetch-x86_64-linux.zip, pinning tox86_64is correct and avoids mixing architectures. No versions or release steps change.Strengths
zigfetch-x86_64-linux.zip), so[x86_64, debian]matches the output instead of relying on a mixed/default pool.snitchuses the same[x86_64, debian]labels).libpci-devviaapt, so the toolchain dependency remains explicit rather than assumed from the image.Issues
Minor
.gitea/workflows/release.yml: the workflow only triggers onv*tags, so the new runner/label combination and the image's tooling assumptions (sudo,zip,curl,jq,node, plus apt access tolibpci-dev) will be exercised for the first time during a real release. If any assumption is wrong, the failure lands on the release itself. A dry run (temporaryworkflow_dispatchor a throwaway pre-release tag) would de-risk this. The author already acknowledged this trade-off, so it is not blocking..gitea/workflows/release.yml:40: the binary is built withzig build -Doptimize=ReleaseSafeand dynamically links libc/libpci (build.zig:32-35). Moving the build host from Ubuntu to Debian trixie (glibc 2.41) can raise the minimum glibc required by the published binary relative to the previous runner. Worth a quick sanity check that the released artifact still runs on the oldest target distro; likely a non-issue if the fleet is on trixie.Verdict
No critical or important issues. This is a safe, minimal infrastructure change. Approve, with the two minor notes above as things to keep in mind for the next release.
APPROVED: true
Reviewed by Botua —
deepseek/deepseek-v4-flash